The Controller Area Network (CAN) protocol is the main communication backbone in modern vehicles, enabling data exchange between Electronic Control Units (ECUs). However, its lack of built-in security mechanisms exposes Connected Vehicles (CV) to a growing range of cyber threats. To address this, we propose a distributed intrusion detection framework that integrates Model-Agnostic Meta-Learning (MAML) with Federated Learning (FL), offering adaptability and privacy preservation. The system leverages an LSTM-based model to learn temporal patterns in CAN message IDs, enabling the detection of known and previously unseen attacks. MAML enhances the model's generalization by facilitating rapid adaptation to new threat types using limited data, while FL enables collaborative training across multiple CVs without sharing raw data. Our distributed approach ensures continuous learning and robustness in real-world automotive environments. Experimental results demonstrate the framework's effectiveness in detecting complex intrusions while maintaining data privacy across participating vehicles.
This paper addresses the joint design of Unmanned Aerial Vehicles (UAVs) trajectory and radio resource management (RRM) in dynamic wireless environments by leveraging a multi-agent deep reinforcement learning (MADRL) framework. In contrast to prior works that either assume constant synchronization between agents and the controller or overlook the communication cost, we explicitly model the interaction between UAVs and the central controller. We propose an adaptive synchronization strategy that selectively transmits model parameters and experience data based on their relevance, enabling a resource-aware RRM algorithm that optimally balances learning performance and communication overhead. The MADRL agents optimize their trajectories based on rewards that incorporate priorities derived from the RRM layer, which jointly manages both uplink and downlink communications. Simulation results demonstrate that our event-driven synchronization strategy outperforms periodic baselines in both convergence speed and communication overhead, towards scalable deployment in realistic urban environments.
Dense wireless networks, such as electromagnetic nanonetworks, are characterized by high number of resource-constrained nanonodes close to each other. Testing communication protocols for these networks is an important challenge as real-world experimentation is complex and theoretical analysis is often too limited. Therefore, simulations are widely used as an alternative due to its cost-effectiveness, reproducibility, and efficiency. However, their accuracy depends on how they simulate the real conditions of the network. For instance, simulation accuracy depends widely on the packet loss model used. This paper addresses this challenge, by comparing the Unit Disc Graph (UDG) and shadowing packet loss models in the context of flooding protocols. The obtained results show that while both models can ensure similar outcomes in some cases, they diverge significantly in others, particularly when data delivery is not guaranteed across the entire network. Finally, this study shows the importance of selecting appropriate packet loss models in simulations to ensure reliable and generalizable protocol evaluation.
Non-Terrestrial Networks (NTNs), comprising Unmanned Aerial Vehicles (UAVs) and High Altitude Platform Stations (HAPS) equipped with Mobile Edge Computing (MEC), offer promising solutions for network traffic and tasks offloading from ground users. To enhance the reliability and energy efficiency of such systems, Reconfigurable Intelligent Surfaces (RIS) can be deployed to control wireless signal propagation. In this context, we propose in this paper a novel MEC-enabled framework with HAPS and RIS-equipped UAVs (UAV-RISs) to optimize task offloading from ground users. Our objective is to minimize the tasks' average end-to-end (E2E) delay, under constraints of UAV and HAPS power capacity and E2E service delay threshold, through the optimization of task assignment and UAV-RIS phase-shift configuration. Given the NP-hardness of the problem, we decompose it into two subproblems. The first consists of optimizing the RIS phase shifts to minimize the RIS-assisted communication delay. The second tackles the task assignment problem using a Particle Swarm Optimization (PSO)-based approach, considering the RIS phase shifting solution previously developed. Through simulations, we validate the efficacy of our approach, which outperforms other benchmarks in terms of task average E2E delay and task offloading success rate.
DECT 2020 NR is a Radio Interface Technology (RIT) introduced by ETSI for massive Machine Type Communications (mMTC) and Ultra-Reliable Low Latency Communications (URLLC) use cases. Notably, DECT 2020 NR stands as the sole non-cellular 5G technology apt for industrial IoT (IIoT) scenarios. However, achieving the stringent reliability and latency requirements of URLLC in dense industrial settings poses a significant challenge due to various propagation losses and severe shadowing. In this research work, we propose a novel DECT 2020 NR framework incorporating Intelligent Reflecting Surface (IRS) to meet the challenging URLLC constraints. The framework maps the system model in 3D space, including the 3D mapping of the radiation pattern of IRS, and employs the Separating Axis Theorem (SAT) to systematically identify shadowed regions within this 3D system model and mitigate the identified shadowed areas, leveraging the proposed novel MAC procedure for the IRS-assisted DECT 2020 NR network. Additionally, we propose a resource allocation scheme based on the priority given to radio devices ready to offload the data. Comparative analysis demonstrates that our proposed solution achieves significant improvements in received power, throughput, and reduction in round-trip time (RTT) when compared to both the existing approach and the conventional DECT 2020 NR standard.
This paper investigates the impact of Line-of-Sight (LoS) path misalignment in distributed Multiple-Input Single-Output (MISO) systems, particularly in millimeter-wave (mmWave) bands where phase synchronization and timing precision are critical. The goal is to validate a bit error rate model under practical OFDM transmission scenarios involving multiple access points communicating with user equipment. We highlight how time misalignment between LoS paths can degrade system performance and evaluate the effectiveness of various compensation strategies under such conditions. A key contribution of this work is the proposed compensation technique for LoS path misalignment. By applying a frequency-dependent linear phase shift to each subcarrier—an operation that is both low in complexity and easy to implement—we induce a circular rotation of the OFDM time-domain symbol. This rotation effectively aligns the various LoS contributions in time, as long as the path delays remain within the cyclic prefix duration. Notably, the method is fully compatible with analog beamforming architectures, which are essential in mmWave systems to reduce the number of costly RF chains. As a result, the proposed strategy enables efficient time compensation with minimal computational and hardware overhead, making it particularly attractive for distributed Multiple-Input Multiple-Output (MIMO) deployments at mmWave frequencies.
Unmanned Aerial Vehicles (UAVs) have seen extensive growth in both civilian and military sectors, driven by their versatility, ease of deployment, and expanding range of applications. However, this expansion has also exposed UAV systems to a growing number of cybersecurity threats. The MAVLink protocol, widely adopted for UAV-ground station communication, lacks native encryption, making it susceptible to interception and manipulation. This paper proposes a comprehensive solution based on the use of Hardware Security Modules (HSMs) to protect cryptographic keys and secure data transmission using the ChaCha20 encryption algorithm. Our approach integrates the HSM into the drone system, ensuring that sensitive keys are securely generated, stored, and accessed. We present the architecture, implementation details, and simulation results that confirm the feasibility and efficiency of the proposed solution, highlighting its potential to significantly enhance the security of UAV communications with minimal performance overhead.
We propose Privacy-Enhanced Secure Neighbor Discovery (PE-SND), a lightweight protocol that addresses the security and privacy needs of existing and emerging wireless networks. Traditional Secure Neighbor Discovery (SND) protocols effectively mitigate relay attacks but expose device identities and locations. Our protocol preserves SND security guarantees while enhancing privacy through pseudonymous authentication and encrypted location exchange. Formal verification confirms pseudonym unlinkability and location confidentiality against both external adversaries and honest-but-curious participants. Performance evaluation across security levels (96 to 256 bits) demonstrates feasibility with processing latencies of 14 to 75ms. Deployment of PE-SND using Ultra-Wide-Band (UWB) technology confirms sub-meter accuracy (15cm) in indoor Line-of-Sight (LoS) and 0.953m in Non-Line-of-Sight (NLoS) environments, with execution times up to 90ms. PE-SND provides a lightweight solution for privacy-demanding applications in the Internet of Things (IoT), Internet of Vehicles (IoV), and other emerging wireless ecosystems.
Nowadays, Wireless Multimedia Sensor Networks (WMSNs) plays a vital role in modern surveillance systems and enables innovative solutions across diverse sectors. In such networks, detecting and minimizing data redundancies has become crucial for both prolonging network lifespan and improving data quality. In this paper, we propose a two-layer reduction scheme, called as SHOLO, that efficiently detects and removes similarities in SHOrt and LOng term video data collected in WMSNs, thus conserve sensor energies and extending network lifetime. In the first layer, SHOLO detects and removes similarities among successive frames captured by video node during the same period time through two novel distance-based methods: Difference Hash (DH) and Pixel Intensity Threshold (PIT). In the second layer, SHOLO searches long-term similarities among frames collected in the same period to detect scene variation and zone dynamicity. Then, we introduced two similarity reduction methods in such layer: Intensity-based Difference (ID) and Block-based Difference (BD). We conducted extensive simulations using real-world video data set to validate the efficiency of the proposed framework. The results demonstrated that SHOLO can reduce up to 93.8% of collected video data, leading to tremendous energy savings and network lifetime compared to existing approaches.
The proliferation of IoT devices, amplified by 6G, has heightened security risks. To counter these, experts are improving Intrusion Detection Systems (IDSs) using Machine Learning (ML) and Deep Learning (DL) algorithms, which rely on Feature Selection (FS) to identify key features and optimize performance in detecting attacks. In this context, this paper proposes a novel informed FS technique that exploits XAI (eXplainable Artificial Intelligence). The proposed algorithm extends the use of Shapley Value in XAI by enhancing it with an accuracy-driven binary search with the aim of finding the most representative feature and therefore reducing the dimensionality of a Network Intrusion Dataset, improving its detection ability. Through a comprehensive experimental campaign, the proposal has been validated and its benefits, which are not only limited to reduced training and testing time, but also drastically streamlining the AI model complexity. Finally, a comparison with other standard FS techniques is also provided to further highlight the advantages of the proposed algorithm. The proposal has been validated through a comprehensive experimental campaign in which results show the benefits. The experimental section shows the performance of the model when trained on the set of features returned by DeepSHAP and on the set returned by our framework, showing a drastic reduction in training and testing time.
The increasing complexity of modern WiFi networks aligns them more closely with cellular systems. This convergence underscores the need for WiFi-specific LLMs, akin to ongoing efforts in 5 G. An essential initial step is the design of a WiFi dataset compatible with LLM requirements, structurally coherent, and containing both technical and general information to ensure broad applicability. This work introduces WiFiQnA, a curated dataset of WiFi-related multiple-choice questions designed for LLM fine-tuning. We define two multiple-choice question (MCQ) formats: general knowledge and procedural configuration/troubleshooting questions. We develop a multi-step generation framework using three LLMs for question generation and four for validation. The process integrates filtered telecom datasets, WiFi-specific sources, and tailored prompts, ensuring semantic diversity, accuracy, and reliability.
The blockchain oracle problem is a central challenge in the integration of blockchain in decentralized systems. Ensuring that off-chain data fed into smart contracts is reliable is a problem, and relying on a single oracle introduces a single point of failure. To address this, several decentralized oracle designs have been proposed, including those based on threshold signature schemes. In such systems, a data feed is accepted only if a minimum number of oracles sign it. While this improves robustness, it introduces coordination issues: signing incurs a cost, and individual oracles may prefer to free-ride, expecting others to sign. In this work, we model oracle participation as a discrete public goods game and analyze the conditions under which signing is a rational strategy in equilibrium. We characterize the set of pure and symmetric mixed-strategy Nash equilibria and study how key system parameters, such as the number of oracles, the cost-to-reward ratio, and the signature threshold, affect participation incentives. Our results show that system parameters can give rise to multiple symmetric mixedstrategy equilibria, but that such equilibria disappear when the signing cost reaches as little as 27.5% of the reward.
Low-Power Wide Area Networks (LPWANs), such as LoRaWAN, are pivotal for large-scale IoT deployments. However, traditional stationary gateways (GWs) impose scalability and cost constraints. We propose an AI-driven mobile GW architecture that leverages reinforcement learning (RL) to dynamically adapt GW mobility based on real-time network conditions. Our custom mobility module, integrated into OMNeT++ with FLoRa and TensorFlow Lite Micro, enables on-the-fly decision-making to optimize Packet Delivery Ratio (PDR) and fairness. The results of the simulations show that the introduced RL-GW achieves up to 99.94 % PDR in high-fidelity simulations, outperforming static and heuristic mobile strategies. The RL policy generalizes robustly across network sizes and scenarios, offering a scalable, low-overhead solution for adaptive LPWAN infrastructure.
Virtual Network Embedding (VNE) approaches typically assume static or slowly-changing network topologies, but emerging applications require deployment in mobile environments where traditional methods become insufficient. This work extends VNE to constrained mesh networks of mobile edge devices, addressing the unique challenges of rapid topology changes and limited resources. We develop models incorporating device capabilities, connectivity, mobility and energy constraints to evaluate optimal deployment strategies for mobile edge environments. Our approach handles the dynamic nature of mobile networks through three allocation strategies: an integer linear program for optimal allocation, a greedy heuristic for immediate deployment, and a multi-objective genetic algorithm for balanced optimization. Our initial evaluation analyzes application acceptance rates, resource utilization, and latency performance under resource limitations. Results demonstrate improvements over traditional approaches, providing a foundation for VNE deployment in highly mobile environments.
This paper addresses the spectrum sensing challenges in Cognitive Radio-based Internet of Things (CR-IoT) networks, which are characterized by intermittent primary user (PU) activities. Unlike conventional approaches relying on fixed detection thresholds, we propose and evaluate an adaptive thresholding method within an Energy Detection (ED) framework, dynamically optimized based on real-time noise and signal characteristics. The PU activity is realistically modeled as a two-state Markov chain, with signals transmitted via Binary Phase Shift Keying (BPSK) modulation over an Additive White Gaussian Noise (AWGN) channel. Simulation results demonstrate that our adaptive threshold approach significantly enhances detection accuracy while reducing total error probability at low SNR conditions (from -25 dB to 0 dB), thus effectively increasing idle channel utilization compared to traditional fixed threshold techniques.
Although the performance of mobile devices (MDs) has been steadily improving, advanced applications might suffer long latency for complex functions as well as from draining the battery when continuously used. Function offloading to cloud servers is a proposed technique to solve these challenges. Recently edge computing enables low latency access to resources and, in combination with Function-as-a-service (FaaS), is an obvious target for offloading from MDs. In this paper we introduce a framework for FaaS applications on MDs and function offloading to a FaaS Edge implementation. The scheduling of function invocations to local or remote resources is determined by function-specific offloading policies. The policies are selected periodically according to historic data and function specific performance and energy models. The periodic planning considers the distribution of the application's energy budget across a number of periods and adapts the functionspecific policies for a period. This enables more time consuming planning and fast decision making for individual function invocations. Our results validate the efficiency of the proposed approach, showing about 33 % reduction in energy consumption.
Spectrum sensing at terahertz (THz) frequencies presents significant challenges due to extreme signal attenuation, device heterogeneity, and real-time processing constraints. In this work, we propose a lightweight federated learning (FL) framework for cooperative spectrum sensing, tailored to decentralized secondary users operating under non-identically distributed signal conditions. We adopt a lightweight convolutional neural network (CNN) with a multi-head attention refinement and soft attention-based pooling, enabling efficient processing of magnitude-domain MIMO-OFDM signals while preserving key spectral features. We evaluate our attention-enhanced CNN model within a federated learning setup, comparing it against a standard FL aggregation baseline, a centralized model trained on aggregated data, and a local-only model trained on a single device. Results show that our model with FedProx consistently outperforms all alternatives, highlighting the benefits of proximal regularization under non-IID conditions, and demonstrating that federated learning can match centralized performance while preserving privacy and scalability in heterogeneous THz environments.
In distributed multi-robot exploration, effective communication among mobile robots significantly impacts the efficiency of area coverage, mapping accuracy, and collaborative decision-making. Typically, robots share their information map based on dynamically established network topologies to balance communication overhead and information accuracy. This paper comparatively analyzes three dynamic topology construction strategies for mobile robots exploring unknown environments: the Relative Neighborhood Graph (RNG), the k-Relative Neighborhood Graph (K-RNG), and the K-Nearest Neighbors (KNN). RNG is a graph-based approach that maintains adaptive links based on geometric proximity, efficiently managing connectivity while reducing redundant data exchanges. The K-RNG generalizes RNG by allowing a tunable number of points within the geometric neighborhood, providing flexible control over network density. In contrast, KNN selects neighbors by prioritizing robots in immediate proximity, quickly adapting to local density changes. We conduct extensive simulations to evaluate these strategies against critical Quality-of-Service (QoS) metrics, including packet delivery ratio and latency. Results demonstrate RNG's effectiveness in reducing unnecessary data transmissions while maintaining stable connections, K-RNG's performance tunability based on the parameter $k$, and the strong dependency of KNN's performance on the choice of $k$.
As artificial intelligence and machine learning (AI/ML) models become integral to network operations, their lack of transparency poses a significant barrier to operator trust. Existing explainable artificial intelligence (XAI) techniques often fail to bridge this gap for non-specialists, producing technical outputs that are difficult to translate into actionable insights. This paper presents a framework specifically designed to address this shortcoming. It leverages a moderately sized large language model (LLM) and extends beyond the standard use of SHapley Additive exPlanations (SHAP) feature influence values. The framework employs a structured prompt enriched with mutual feature interaction data to generate human-understandable natural language explanations. To validate our framework, we performed an empirical evaluation on an optical quality of transmission (QoT) estimation use case with human evaluators. We collected independent performance evaluations from specialists, which showed a high inter-evaluator agreement. Compared to a state-of-the-art baseline that uses only SHAP feature influence values in a straightforward prompt, our approach improves the explanation usefulness and scope by 12.2% and 6.2%, while achieving 97.5% correctness.
As chronic diseases continue to rise globally, public health systems are under increasing pressure to find innovative and scalable care solutions. In response, this study introduces an AI-powered health advisory system that leverages generative AI techniques, specifically retrieval-augmented generation (RAG) and integrates seamlessly with LINE and the ChatGPT API. Designed to support individuals managing chronic conditions, the system provides real-time, personalized recommendations, including dietary guidance, medication reminders, and interpretations of health checkup results. Experimental evaluations show that the system achieves over 90% accuracy across key functions, underscoring its potential to enhance self-management and support preventive healthcare strategies.
